N early 20 years ago, the US National ScienceFoundation (NSF) convened a panel to report on

the potential of visualization as a new technology.1 Lastyear, the NSF and US National Institutes of Health(NIH) convened the Visualization Research Challenges(VRC) Executive Committee—which was made up ofthe authors of this article—to write a new report. Here,we summarize that new VRC report, available in full athttp://tab.computer.org/vgtc/vrc and in print.2

The goals of this new report are to

■ evaluate the progress of the maturing field of visual-ization,

■ help focus and direct future research projects, and■ provide guidance on how to apportion national

resources as research challenges change rapidly in thefast-paced world of information technology.

We explore the state of the field, examine the potentialimpact of visualization on areas of national and inter-national importance, and present our findings and rec-ommendations for the future of our growing discipline.Our audience is twofold: the supporters, sponsors, andapplication users of visualization research on the onehand, and researchers and practitioners in visualizationon the other. We direct our discussion toward solvingkey problems of national interest and helping this work’ssponsors to concentrate resources to the greatest effect.Our findings and recommendations reflect informationgathered from visualization and applications scientistsduring two workshops on VRC, as well as input from thelarger visualization community.

The value of visualizationAdvances in the science and technology of comput-

ing have engendered unprecedented improvements inscientific, biomedical, and engineering research;defense and national security; and industrial innova-tion. Continuing and accelerating these advancementswill require people to comprehend vast amounts of dataand information produced from a multitude of sources.3

Visualization, namely helping people explore or explaindata through software systems that provide a static orinteractive visual representation, will be critical inachieving this goal. Visualization designers exploit thehigh-bandwidth channel of human visual perception to

help people comprehend information orders of magni-tude more quickly than through reading raw numbers ortext.

Visualization is fundamental to understanding mod-els of complex phenomena, such as multilevel modelsof human physiology from DNA to whole organs, mul-ticentury climate shifts, international financial markets,or multidimensional simulations of airflow past a jetwing. Visualization reduces and refines data streams,thus enabling us to winnow huge volumes of data inapplications such as public health surveillance at aregional or national level to track the spread of infec-tious diseases. Visualizations of such application prob-lems as hurricane dynamics and biomedical imaging aregenerating new knowledge that crosses traditional dis-ciplinary boundaries. Visualization can provide indus-try with a competitive edge by transforming businessand engineering practices.

Although well-designed visualizations have thepower to help people enormously, naive visualizationattempts are all too often ineffective or even activelymisleading. Designing effective visualizations is a com-plex process that requires a sophisticated understandingof human information-processing capabilities, bothvisual and cognitive, and a solid grounding in the con-siderable, already existing body of work in the visual-ization field. Further research in visualization—and thetechnology transfer of effective visualization method-ologies into the working practice of medicine, science,engineering, and business—will be critical in handlingthe ongoing information explosion. The insights pro-vided through visualization will help specialists discoveror create new theories, techniques, and methods, andimprove the daily lives of the general public.

While visualization is itself a discipline, advances invisualization lead inevitably to advances in other disci-plines. Just as knowledge of mathematics and statisticshas become indispensable in subjects as diverse as thetraditional sciences, economics, security, medicine, soci-ology, and public policy, so too is visualization becom-ing indispensable in enabling researchers in other fieldsto achieve their goals. Like statistics, visualization is con-cerned with the analysis and interpretation of informa-tion, both quantitative and qualitative, and with thepresentation of data in a way that conveys their salientfeatures most clearly. Both fields develop, understand,

TamaraMunznerUniversity ofBritishColumbia

Chris JohnsonUniversity ofUtah

RobertMoorheadMississippi StateUniversity

HanspeterPfisterMitsubishiElectric ResearchLaboratories

PennyRheingansUniversity ofMaryland,BaltimoreCounty

Terry S. YooNational Libraryof Medicine

NIH-NSF Visualization Research Challenges ReportSummary __________________________________________

Visualization Viewpoints

Editor: Theresa-Marie Rhyne

20 March/April 2006 Published by the IEEE Computer Society 0272-1716/06/$20.00 © 2006 IEEE

and abstract data-analytic ideas and package them inthe form of techniques, algorithms, and software for amultitude of application areas.

However, despite the importance of visualization todiscovery, security, and competitiveness, support forresearch and development in this critical, multidiscipli-nary field has been inadequate. Unless we recommit our-selves to substantial support for visualization research,development, and technology transfer, we will see adecline in the progress of discovery in other importantdisciplines dependent on visualization. As these disci-plines lose their ability to harness and make sense ofinformation, the rate of discovery itself will decline. Inthe inevitable chain reaction, we will lose our competi-tive edge in business and industry.

FindingsVisualization is indispensable to the solution of com-

plex problems in every sector, from traditional medical,science, and engineering domains to such key areas asfinancial markets, national security, and public health.Advances in visualization enable researchers to analyzeand understand unprecedented amounts of experi-mental, simulated, and observational data and throughthis understanding to address problems previouslydeemed intractable or beyond imagination. Yet, despitethe great opportunities created by and needs fulfilledby visualization, the NSF and NIH (and other US gov-ernment agencies) have not effectively recognized thestrategic significance and importance of visualizationin either their organizational structures or their researchand educational planning. The recent and cogent bookoutlining the visual analytics research agenda,4 whichdovetails closely with visualization, offers promise thatvisualization for the national security application areawill be well funded in the near future. However, thateffort encompasses only one sector and is short term,whereas the field needs long-term support across manysectors. The current distribution of funding sources doesnot reflect the potential benefits of visualization researchto specific application areas. These inadequacies com-promise the future of US scientific leadership, publichealth, and economic competitiveness.

Many important challenges in visualization will not besolved by simply waiting for computers to get faster, orby refining existing algorithms. Visualization researchersshould collaborate closely with domain experts who haveappropriate driving tasks in data-rich fields to producetools and techniques that solve clear real-world needs.Visualization researchers also need to integrate with andbe informed by methodologies in many other fields—including statistics, data mining, and cognitive psychol-ogy—to facilitate analysis from both qualitative andquantitative perspectives. Examining how and why visu-alizations work will require a deep understanding ofhuman characteristics, strengths, and limitations. The setof current visualization techniques is rich and powerful,but far from complete; we need to systematically explorethe design space of possible visual representations.Designing appropriate interaction metaphors for currentand future hardware will be crucial for harnessing thefull power of visualization systems.

Evaluating the extent to which proposed visualiza-tion techniques solve problems grounded in real-worldtasks is critical to the success of our field. Quantitativeevaluation methods include measurement of algorithmexecution times and memory performance, and formaluser studies in a laboratory setting aimed at generatingstatistically significant results measuring user perfor-mance on abstracted tasks using metrics such as taskcompletion times or error rates. Qualitative evaluationapproaches include anecdotal evidence of discoveriesby real-world users, adoption rates based on user-com-munity size, and qualitative user studies ranging fromethnographic analysis of target-user work practices tolongitudinal field studies to informal usability evalua-tion of a prototype system. Finally, evaluative analysiswith careful justification of design choices with respectto conceptual and theoretical frameworks underlying

IEEE Computer Graphics and Applications 21

Teaching Anatomy and Planning Surgery Detailed anatomic models of delicate organs such as the human

hand are a prerequisite for both teaching the complex anatomy andthe preoperative simulation of interventions. While classicalanatomy atlases can provide sufficient anatomical detail in a set ofstatic images, they do not allow choosing user-defined views orperforming surgical interaction. The picture and inset in Figure Aillustrates Voxel-Man/Upper Limb, a novel computer-basedanatomy model that not only allows arbitrary viewing anddissection, but also the interrogation of the anatomic constituentsby mouse click. The pictorial model was created from the VisibleHuman Project data set using a combination of volume visualizationfor bone and muscles, and surface modeling for blood vessels,nerves, ligaments, and tendons.1 The pictorial model is linked to aknowledge base, describing the anatomic constituents and theirrelations. Voxel-Man supports a range of practitioners, from medicalstudents to expert surgeons, in coping with the complexity of state-of-the-art microsurgical interventions. A future challenge will be toextend the knowledge base so that the system can warn the user ofconsequences of a surgical interaction for the patient.

A View of Voxel-Man/Upper Limb. (Reprinted with permission ofWiley-Liss. a subsidiary of John Wiley & Sons.)

Reference1. S. Gehrmann et al., “A Novel Interactive Anatomic Atlas of the Hand,”

Clinical Anatomy, DOI 10.1002/ca.20266, 2006.

the visualization field is useful both for understandingwhy and how a method works, and for extending theframeworks themselves.

One of the basic requirements of science is that exper-iments be repeatable. There is great need for sustainedand methodical creation and maintenance of curateddata, model, and task repositories so that visualizationresearchers can benchmark new algorithms and tech-

niques by comparing them directly to the results of pre-vious work. Although visualization practitioners are typ-ically not the primary source of the data themselves,they should be advocates for open science through datasharing whenever possible.

In visualization, as in other research areas, researchcan be divided into the three categories of basic or foun-dational work, transitional approaches to create andrefine techniques, and application-driven efforts.Although transitional research is at the heart of anyfield, in some areas of visualization a disproportionateamount of attention is currently devoted to incremen-tal refinement of a narrow set of techniques. Cycles ofinteraction between all three of these areas must occurto keep the field of visualization vibrant, driven byapplied problems and grounded in foundationalresearch. Moreover, just as being engaged with domainscientists will help visualization research become moreeffective, being engaged with visualization researcherswill help domain scientists by providing them withaccess to state-of-the-art tools. These collaborations willsupport grass-roots technology transfer of visualizationmethodologies into application domains in the eco-nomic, scientific, and public service sectors.

Revisiting the hardware, software, and networkingissues discussed in the 1987 report reveals the enor-mous progress made in the past decades. Fast andcheap commodity hardware meets most of the CPUand graphics needs of visualization today. The currentavailability of commercial volume-rendering hardwareis a success story for the field. Exciting new advancesin display hardware will have a major impact on visu-alization. Advances in networking will greatly facili-tate collaboration and other remote visualizationopportunities. Both commercial and open source visu-alization software systems are thriving as the fieldmatures. The open source model offers many benefitsto both academia and industry, and to both researchersand end users.

RecommendationsWe explored these findings with the assistance of area

experts in two separate workshops. Based on the delib-erations of our panelists, we make the following rec-ommendations.

Principal agency leadership recommendationNSF and NIH must make coordinated investments

in visualization to address the 21st century’s mostimportant problems, which are predominantly collab-orative, crossing disciplines, agencies, and sectors.Both NSF and NIH can and should provide leadershipto other federal funding partners and to the researchcommunities they support. To achieve this, NSF andNIH should modify their programmatic practices tobetter engage visualization capabilities across disci-plines important to scientific and social progress and toencourage and reward interdisciplinary research, openpractices, and reproducibility in technical and scientif-ic developments. Such agency leadership is critical ifwe are to meet US needs in critical areas and to main-tain US competitiveness in a global environment.

Visualization Viewpoints

22 March/April 2006

Characterizing Flow Visualization MethodsFor decades, researchers have been developing visualization

techniques that advance the state of the art and are published inpeer-reviewed journals. However, there are disproportionately fewquantitative studies comparing visualization techniques, such asthe characterization in Figure A of the differences between flowvisualization methods—showing six methods for visualizing thesame 2D vector field.1 Subjects who participated in the user studyperformed several tasks including identifying the type and locationof critical points in visualizations. Assuming roughly equalimportance for all tasks, the streamline visualization shown inFigure A6 performed best overall: on average, subjects were fastestand most accurate when using it.

A A comparison of six methods for visualizing the same 2D vectorfield: (1) icons on a regular grid, (2) icons on a jittered grid, (3) layer-ing method inspired by oil painting, (4) line-integral convolution, (5)image-guided streamlines, (6) streamlines seeded on a regular grid.(Courtesy of David Laidlaw, Brown University.)

This study produced both quantitative results and a basis forcomparing other visualization methods, for creating more effectivemethods, and for defining additional tasks to further understandtradeoffs among methods. A future challenge is to developevaluation methods for more complex 3D time-varying flows.

Reference1. D.H. Laidlaw et al., “Comparing 2D Vector Field Visualization Methods:

A User Study,” IEEE Trans. Visualization and Computer Graphics, vol. 11, no. 1, 2005, pp. 59-70.

(1) (2) (3)

(4) (5) (6)

Short-term policy recommendationPolicy changes for both funding and publication

review to encourage evaluation of visualization and col-laboration between visualization and other fields canbe implemented immediately, without requiring newfunding initiatives. Within visualization, review proto-cols should reflect the importance of evaluation to deter-mine the success and characterize the suitability oftechniques. Methodological rigor should be expectedwhen user studies are proposed or reviewed, and as nec-

essary, visualization researchers should collaborate withthose trained in fields such as human–computer inter-action, psychology, and statistics. Peer review of pro-posals and publications should also reward visualizationdriven by real-world data and tasks, in close collabora-tion with target users. In other fields, review protocolsshould encourage domain scientists who create part-nerships with visualization researchers, just as engage-ment with statisticians is considered normal practice inmany areas such as biomedical research.

IEEE Computer Graphics and Applications 23

The Visualization Toolkit In 1993, three visualization researchers from

the General Electric corporate R&D center beganto develop an open source visualization system.This system, which came to be known as theVisualization Toolkit (VTK), was initiallyenvisioned as a teaching and researchcollaboration tool, hence its release under anopen source license. The software gained rapidacceptance, in part due to the sophistication ofits object-oriented design and software process,but also because of the community of users thatformed around it. VTK is now in worldwide use,and has helped spawn several small companiesand derivative products. For example, Kitware,Inc. was formed in 1998 to support VTK,subsequently creating products based on thetoolkit including the open source ParaViewparallel visualization system and the proprietaryVolView volume-rendering application.1 VTKcontinues to evolve with contributions fromresearchers in academia, US nationallaboratories, and businesses, and is used indozens of commercial software applications.

Figure A shows a computational fluiddynamics visualization using ParaView (FigureA1) and volume rendering using VolView (FigureA2). Figure A3 uses the LOx Post data set,simulating the flow of liquid oxygen across a flatplate with a cylindrical post perpendicular to theflow. This model studies the flow in a rocketengine, where the post promotes the mixing ofthe liquid oxygen.2

References1. W.J. Schroeder, K. Martin, and B. Lorensen, The

Visualization Toolkit: An Object Oriented Approach toComputer Graphics, 3rd ed., Kitware, Inc., 2004.

2. S.E. Rogers, D. Kwak, and U.K. Kaul, “A Numerical Study of Three-Dimensional Incompressible Flow around Multiple Posts,” Proc. AIAA Aerospace Sciences Conf., AIAA paper 86-0353, 1986, pp. 1-12.

A Sample visualizations from publicly availablevisualization software: (1) ParaView, (2) VolView,and (3) VTK. (Courtesy of Kitware Inc.)

(1)

(2)

(3)

Midterm direction recommendationPilot programs should be established to combine

efforts and create collaborative development betweenvisualization and other research domains. Funding forsuch programs should contribute proportionately to boththe visualization research and the domain specialty. Thepurpose of this effort will be to improve the penetrationof emerging technologies into new domains, increasingtheir facility to move data and share results through visu-alization. All of the awards in this area should be dedi-cated to open access of source code, availability ofresearch data to the worldwide community, and repro-ducibility of the technical and scientific developments.

Long-term investment recommendationWe recommend a coordinated and sustained nation-

al investment in a spectrum of centralized and distrib-uted research programs to promote foundational,transitional, and applied visualization research in sup-port of science, medicine, business, and other sociallyimportant concerns. This investment is critical for theUS to remain competitive in a global research and devel-opment community that has increasing resources. Inaddition to funding transitional research, such programsshould emphasize foundational research and integra-tion of methodologies from other fields, and collabora-tion with domain specialists who provide drivingproblems in areas of national concern. A long-term fund-ing commitment is required for the creation and main-tenance of curated data collections, and open sourcesoftware, to promote open science. Characterizing howand why visualizations work, systematically exploringthe design space of visual representations, developingnew interaction approaches, and exploiting the possi-bilities of novel display hardware will be particularlyimportant areas of emphasis. ■

AcknowledgmentsWe are indebted to the expert and visionary VRC

workshop panelists for sharing their considerable tal-ents and insights, and to NIH and NSF for their spon-sorship. In particular, we thank panelists ManeeshAgrawala, Liz Bullitt, Steve Cutchin, David Ebert,Thomas Ertl, Steve Feiner, Felice Frankel, Bob Galloway,Alyssa Goodman, Mike Halle, Pat Hanrahan, ChuckHansen, Helwig Hauser, Karl Heinz Höhne, Ken Joy,Arie Kaufman, Daniel Keim, David Laidlaw, Ming Lin,Leslie Loew, Bill Lorensen, Wayne Loschen, PatrickLynch, Kwan-Liu Ma, Alan MacEachren, Chris North,Art Olson, Catherine Plaisant, Jerry Prince, WillSchroeder, Jack Snoeyink, John Stasko, BarbaraTversky, Matthew O. Ward, Colin Ware, Ross Whitaker,Turner Whitted, and Jarke van Wijk.

References1. B.H. McCormick, T.A. DeFanti, and M.D. Brown, Visual-

ization in Scientific Computing, National Science Founda-tion, 1987.

2. C. Johnson et al., NIH-NSF Visualization Research Chal-lenges Report, IEEE Press, 2006.

3. P. Lyman and H.R. Varian, How Much Information, 2003;http://www.sims.berkeley.edu/how-much-info-2003.

4. J.J. Thomas and K.A. Cook, eds., Illuminating the Path: TheResearch and Development Agenda for Visual Analytics,National Visualization and Analytics Center, 2005.

Contact Terry Yoo at tyoo@mail.nih.gov.

Contact editor Theresa-Marie Rhyne at tmrhyne@ncsu.edu.

Visualization Viewpoints

By Michael Blaha Modelsoft Consulting

Applications with fixed data structures can easily deal with a direct database representation. This approach doesn't work well when the structure is not fully known beforehand. This ReadyNote presents softcoded values, a generic mechanism use for defining and storing data at runtime. $19www.computer.org/ReadyNotes

Designing and Implementing Softcoded Values

IEEE ReadyNotes